Autonomous vehicles validation with SOTIF

Autonomous vehicles validation with SOTIF


Hi My name is Fabrice TROLLET, and I am the MaTeLo product Manager at All4tec. Today, you will learn to create amazing test cases for autonomous vehicle validation using simulation. I am going to show you an easy industrial
method where you”ll be able to generate and execute complex test cases by yourself. Starting from your requirements, we provide maneuver libraries to manage functional unitary tests like those from EuroNCap. For example cut-in,
cut-out, zip-in, merging etc. It is a good starting kit for you building
your own test cases. MaTeLo generates for each unit test maneuver, hundreds of thousands combinatorics. Test are graphic, reproductible, and can be
stored in any ALM or database. Test Cases can be provided in various file
formats, and are executed in 3D simulation tools. Doing the variability of known unitary test
case is compliant with ISO 26 26-2 and the §10 of the SOTIF Verification
: Evaluate Known use cases. MaTeLo can also execute and schedule maneuver
unit test cases directly into the simulator, with the “online generation
method” within a closed loop between MaTeLo and the simulation tool. This method lets create unknown use cases
as required by the §11 of the SOTIF Validation. MaTeLo advanced algorithms will provide you
a representative sample of variability and combinatorics with minimum
redundancies, to simulate many kilometers, and reduce simulation time. Let’s see the benefits of that method! Explore all the possibilities of your map,
and do sequentially on the same road, the maneuvers described in unit
tests, with all the possible variabilities offered by environment. Merge automatically unit maneuvers in one
complex test case where possible maneuvers are run simultaneously,
to create for example a cut-in, a cut-out and a zip-in at the same
time. Let the traffic runs to create unpredictable
situations, detect and execute possible tests, depending real time
values of the scene or dynamic detected events, maximizing the number
of test during simulation. Achieve a big variability within all variables
and converge quickly to distribution and get the minimum sample combinatorics with the best test coverage. Calculate the test oracle and check for each test the result to conclude the verdict. Do the play back of the successive or simultaneous unit tests in a deterministic way, with the same variability as the first simulation, to the nearest simulation step. Obtain a report on variability, coverage of
equivalence classes and requirements, and get the detail in graphical
format for each test. Now let see the Online SCANeR example that
is provided as a MaTeLo demo project. This project is ready to use and allows to
perform tests such as cut-in and cut-out simultaneously. You will be able to execute directly this
project by yourself and you will have a ready to use Proof of Concept. This example is the Online mode where test
cases are created during test generation in a loop between Simulation tool and MaTeLo. We use AV Simulation SCANeR for this example. For each possible unit test, MaTeLo asks SCANeR
to detect real time events during simulation with an initial variability set by MaTeLo. For a cut-in event, the trigger is to have
both a vehicle on another lane close to the position of the EGO and its speed higher than
EGO’s. For a cut-out event, the trigger is to have
two vehicles in front of the EGO, with different intravehicular distances. For a cut-in and cut-out simultaneously, the
trigger is both cut-in and cut-out triggers. When events are detected, SCANeR is set to
pause, then MaTeLo gets the data from the scene like speeds, positions, sizes of involved
vehicles in the detected maneuvers. Then MaTeLo chooses parameter variability
to achieve possible maneuvers, like for example the time to go to right lane. Theses parameters are randomly chose depending
physical possibilities of the scene to achieve the maneuvers without impossible behavior. MaTeLo variability algorithm will go quickly
toward distribution of each equivalence classes of parameters with the minimum of doubles,
to generate minimum tests with the highest representativity. Then MaTeLo asks SCANeR to perform maneuvers
with other conditions to detect when maneuvers are finished. When maneuvers are achieved, SCANeR is set
to pause a second time, then MaTeLo gets data of the new scene, and computes the verdict
of the test by comparing expected result with measured data. The test is finished and a new test can start
! To go further, download and install MaTeLo
for free on our website all4tec.net. You will be able to execute Online SCANeR
demonstration. You will be ready to build your tests on your
routes with Matelo. Have a good trip!

Add a Comment

Your email address will not be published. Required fields are marked *